In this study, a representative ice crystal habit of general habit mixture Baum et al. (2014) is assumed in the forward simulation of the cirrus case while for the DCC case, a habit of
aggregated plates Yang et al. (2013) is applied. However, it should be taken into cosideration that assuming a single habit can introduce uncertainties since in reality, the ice crystal habit evolves following the development of the cloud. According to van Diedenhoven et al. (2014) and Sassen and Wang (2008), the variability of ice crystal habit is strongly governed by dynamic and thermodynamic state of the cloud, as well as the available supply of ice particle forming nuclei. The information given by in situ measurements capture at a certain state of the cloud only. Therefore, a sensitivity study is carried out to analyze uncertainties resulting from the assumption of ice crystal habit. According to Platnick et al. (2017), the discrepancy on the retrievals ofτ between assuming two different habits (indexed with "A" and "B") can be approximated by:
τA
τB
≈ 1 − ˜ω0,B· gB 1 − ˜ω0,A · gA
, (5.3)
whereω˜0is the single scattering albedo andg is the single scattering asymmetry parameter.
Since retrievals ofτ commonly uses a scattering wavelength (e.g., λ = 645 nm ) character- ized with a single scattering albedoω˜0 ≈ 0, the expressions on the right hand side of Eq.
5.3 can be simplified to1 − g. Fig. 5.6 shows g and the co-albedo (1 − ˜ω0) of GHM (red) by
Baum et al. (2014), aggregated plates (blue), and aggregated columns (black) by Yang et al. (2013) as a function ofreff calculated at three wavelengths: 645 nm (left), 1240 nm (middle),
and 1640 nm (right). All the habits analyzed for this study are modeled with a high surface roughness (so-called severely roughed surface), in accordance with the information given by the in situ measurements during the campaigns (Järvinen et al., 2016; Voigt et al., 2017). It is found that the aggregated columns have the least magnitude ofg (∼ 0.75) compared to the other habits. Due to smallerg, the incoming solar radiation is scattered more to the backward direction, relative to the forward direction, enhancing the upward radiance measured by the sensor above the cloud. As a result, retrievals ofτ assuming a habit with smallerg will result in a smaller τ , and vice versa. Considering the three habits analyzed here, this leads toτagg.columns<τGHM <τagg.plates.
In Fig. 5.6d-5.6f, it is obvious that each habit has distinct values of co-albedo that vary with reff. To approximate the discrepancies for assuming different habits, Platnick et al.
(2017) and Holz et al. (2016) used the co-albedo value of habit "A" with a predetermined reff and then interpolates this value onto habit "B" to find a reff that gives the same co-
albedo value. By this approach, assuming a habit with a smaller co-albedo will result in a larger retrievedreff, and vice versa. Nevertheless, this approach seems to be inappropriate
because if this is the case, the cloud is then considered to consist of only one particle. In reality, it is known that a cloud is composed of many particles. Additionally, it should be kept in mind that retrievals ofreff are not only influenced by the co-albedo solely, but also
by g. Assuming a habit with a smaller g will result in a larger retrieved reff. Due to a
smallerg, the resulting simulated upward radiance is higher, thus a larger reff which gives
facts, a more comprehensive approach in analyzing the discrepancies on the retrievals of reff should into account both the co-albedo andg.
(a) 645 nm 5 15 25 35 45 reff (µm) 0.70 0.80 0.90 1.00 g GHM plate_10elements column_8elements (b) 1240 nm 5 15 25 35 45 reff (µm) 0.70 0.80 0.90 1.00 g (c) 1640 nm 5 15 25 35 45 reff (µm) 0.70 0.80 0.90 1.00 g (d) 645 nm 5 15 25 35 45 reff (µm) 0.0e+00 1.0e-07 2.0e-07 3.0e-07 co-albedo (e) 1240 nm 5 15 25 35 45 reff (µm) 0.0e+00 2.0e-03 4.0e-03 6.0e-03 8.0e-03 co-albedo (f) 1640 nm 5 15 25 35 45 reff (µm) 0.0e+00 3.0e-02 6.0e-02 9.0e-02 co-albedo
Figure 5.6: Single scattering asymmetry parameter g as a function of reff at λ = 645 nm (a), 1240
nm (b), and 1640 nm (c) for different ice crystal habits: GHM (red) by Baum et al. (2014), aggregated plates (blue), and aggregated columns (black) by Yang et al. (2013) (red). (d), (e), (f) are the co-albedo (1 − ˜ω0) at corresponding wavelengths. (a) 0.0 2.5 5.0 7.5 10.0 τ ci 0.0 2.5 5.0 7.5 10.0 τ ci,ret GHM plate_10elements column_8elements (b) 0.0 15.0 30.0 45.0 60.0 r eff,ci (µm) 0.0 15.0 30.0 45.0 60.0 r eff,ci,ret (µm) GHM plate_10elements column_8elements
Figure 5.7: Comparison of synthetically retrieved τ (a) and reff (b). Synthetic measurements are
generated assuming three habits, GHM (red cross), aggregated plates (blue diamond), and aggreg- ated column (black star). The retrievals are performed by assuming GHM.
Fig. 5.7 shows the comparison of synthetically retrievedτ (a) and reff (b). For this purpose,
synthetic measurements are generated via forward simulations by assuming three different habits, GHM (red cross), aggregated plates (blue diamond), and aggregated columns (black star) which coverτ between 1 and 8, and reff between 10 and 45 µm. For the retrieval ofτ
in Fig. 5.7a, thereff is fixed to 25 µm while for the retrieval ofreff in Fig. 5.7b, theτ is fixed
to 3. All the retrievals are made by assuming GHM and using the combination 1 (645 nm andℜ1640). The result in Fig. 5.7 clearly shows that retrievals ofτ assuming GHM result
in larger retrievedτ when in reality, the ice crystals are composed of aggregated columns. On the other hand, smallerτ are obtained when in reality the ice crystals consist of ag- gregated plates. This condition is profoundly influenced bygGHM >gagg.columns andgGHM
>gagg.plates. The resulting relative difference between the original and retrieved values for
the aggregated columns yields values between16 % and 19 % that increases with τ while for the aggregated plates, it ranges between20 % and 30 % which decreases with τ . The result in Fig. 5.7b yields that retrievals of reff assuming GHM result in largerreff when
in reality, the ice crystals are composed of aggregated plates. Conversely, smallerreff are
obtained if in reality, the ice crystals are comprised of aggregated columns. The resulting relative difference between the original and retrieved values for the aggregated columns is between13 % and 16 % that increases with reff while for the aggregated plates, it ranges
between30 % and 49 % which decreases with reff. The results reveal clearly that the dis-
crepancies on the retrievals ofτ and reff are mainly influenced byg. When the co-albedo
gets more distinct, e.g., between GHM and aggregated plates forreff > 20 µ (see Fig. 5.7f),
the influence of co-albedo on the retrievals ofreffis more pronounced but it is considerably
small compared to the impact ofg.